Study on contact state monitoring in monocrystalline silicon grinding process based on Bayesian information criterion and empirical mode decomposition

Monocrystalline silicon, as a typical hard and brittle difficult-to-machine material, requires precise detection of wheel-workpiece contact during grinding to ensure machining quality and efficiency. To address the demand for high-precision monitoring of contact states, an innovative monitoring approach based on empirical mode decomposition (EMD) and Bayesian information criterion (BIC) slope for high-precision contact-state identification using acoustic emission (AE) signals is proposed. Compared with conventional amplitude threshold methods and multiple AE statistical feature-based approaches, the BIC feature demonstrates superior sensitivity to contact events, outperforming conventional metrics such as RMS and variance. The proposed EMD-BIC slope method achieves early detection within 0.002−0.5 ms, showing significant temporal advantages over threshold-based methods (0.05−0.5 ms earlier) and statistical feature methods (0.002−0.4 ms earlier). By integrating time-frequency analysis with statistical inference, this method significantly improves both the accuracy and timeliness of contact detection. The findings provide a reliable technical solution for real-time monitoring and high-performance manufacturing in ultraprecision silicon grinding processes.